Introduction

Attention deficit/hyperactivity disorder (ADHD) is one of the most common neurodevelopmental disorders in childhood and adolescence, which is mainly characterized by age-inappropriate inattention, hyperactivity and impulsivity ¹. Previous studies had found lag in maturation of intrinsic brain functional connectivity (FC) in ADHD ². Recent studies began to demonstrate that cerebral FC is not constant in a static way over time and there is a temporal variability even in resting state ³. Previous study had evaluated the brain connectivity dynamics in the typical developing children age from 6 to 10 years ⁴. However, whether the time-varying pattens of connectivity in ADHD brain are different from typical developing brain remains unknown. Therefore, we aimed to explore the dynamic changes of whole-brain functional network connectivity from childhood into adolescence (7-14 years) in both ADHD children and typical developing control (TDC) children and try to characterize the differences between them.

Materials and Methods

Participants and MR Data Acquisition
A total of 59 drug-naïve ADHD children and 68 TDC children were recruited in this study. These subjects were divided into 4 age groups (early childhood: age = 7-8 years; late childhood: age = 9-11 years; early adolescence: age = 12-14 years). Diagnosis of ADHD was determined by two experienced clinical psychiatrists according to DSM-5. Resting-state fMRI data of all the participants were acquired in 3.0T Siemens scanner, using a gradient-echo echo-planar imaging sequence with slice thickness = 4 mm, slice gap = 0.2 mm, repetition time = 2000 ms, echo time = 30 ms, flip angle = 90°, matrix size = 64×64, field of view = 192×192 mm2. Preprocessing of fMRI data was conducted in Data Processing Assistant for Resting-State fMRI (DPARSF, http://www.restfmri.net, version 4.5) using an automated pipeline.
Dynamic Functional Network Connectivity Analysis
The dynamic analysis was conducted using the sliding window analysis approach and combined independent component analysis (ICA). The ICA was performed to estimated 30 functional independent components using the GIFT toolbox. The dynamic FNC was conducted to capture the time-varying patterns of functional network connectivity using the temporal dFNC toolbox package. The sliding window size was 22 TRs (44 seconds); sliding in steps of 1 TR, resulting in 168 consecutive windows across the entire scan. Then the k-means clustering was applied to evaluate transient states of FNC, and the number of states was determined using the elbow criterion of the cluster validity index.
Statistical analysis of dynamic connectivity measures
The following dynamic connectivity measures were statistically evaluated: (a) fraction times (the percentage of total time a subject spent in each state); (b) dwell times (the time a subject spent in a state without switching to another one); (c) numbers of transitions (how often a subject changed states). Additionally, we performed three-level one-way ANOVAs to test for differences between the three age groups. Post hoc t-tests of two group comparisons were added in case of significant ANOVA results. The level of significance was set at P < 0.05.

Results

The demographics, clinical and cognitive characteristics were shown in Figure 1.
We identified 15 intrinsic connectivity networks using the ICA after excluding other 15 useless components such as motion artifact, white matter etc. (Figure 2). By using the clustering analysis, four transient states were identified which recurred through scans in ADHD and TDC. As shown in the Figure 3, the percentages of total occurrences of four states were quite different in ADHD and TDC, and in each age group. The state 1 (disconnectivity) and state 2 (mild connected) occurred more frequently in all subjects and existed in all age groups of both ADHD and TDC, while state 3 and state 4 (hypoconnectivity and hyperconnectivity) were observed less frequently. Meanwhile, the patterns of FNC of the state 3 and state 4 were different between each age group.
The results of temporal characteristics showed that as age grew, the fraction times and dwell times in state 3 increased in TDC but decreased in ADHD. In addition, the fraction times and dwell times were lowest in late childhood group of TDC but highest in that of ADHD (Figure 4).

Discussion & Conclusion

This is the first study to demonstrate developmental changes of dynamic functional network connectivity patterns and temporal characteristics in ADHD children. The current study yielded three main findings: Firstly, modularized dynamic states of internetwork disconnectivity occurred most frequently and remained stable as age grew in both ADHD and TDC children. Secondly, occurrence percentages of dynamic states of internetwork hyperconnectivity and hypoconnectivity decreased with age in TDC while remained unchanged in ADHD. Thirdly, the developmental changes of duration of stay of internetwork hyperconnectivity and hypoconnectivity states differed between ADHD and TDC children.